Every other module reads a number off the venue and applies a rule. This one asks gpt-5.6-luna what a question actually resolves at, without showing it the price, and bets only when the answer and the market disagree by 20% or more. Most of the time they agree, and agreement is not a trade — so most of this page is opinions the module formed and did not act on.
No market has been priced by the model yet. Every new binary market gets an estimate; this fills as they arrive.
Every market the model priced, with the argument it gave — including the ones it argued itself out of. A multi-choice market is priced one answer at a time and each answer is its own row, named after the question. On this module the reasoning is the strategy, so a decline without it is the one decline nobody can check.
| market | price | model | gap | outcome |
|---|---|---|---|---|
| Nothing priced yet. | ||||
Nothing judged yet.
| utc | version | market | bet | result | why |
|---|---|---|---|---|---|
| 09-01 09:19 | v82 | Who will win the Brawl Stars challengers? :: Fut Freezone | YES 8 | settled +25 | model 0.50; edge +0.42; luna-medium price-blind says 0.50 vs answer 0.08 (edge +0.42 >= 0.1): The market is materially ambiguous because it asks who will win but defines no YES outcome or candidate, so despite a 1/12 base rate for any unspecified team, the estimate is pulled to 0.5. |
| 09-01 09:19 | v82 | Who will win the Brawl Stars challengers? :: Reventant Xspark | YES 3 | settled -3 | model 0.22; edge +0.14; luna-medium price-blind says 0.22 vs answer 0.08 (edge +0.14 >= 0.1): Base rate is 8.3% (1 winner among 12 teams); the rules clearly identify Revenant XSpark despite the typo, and its recent regional title plus top-tier analyst ranking raise it to about 22%, with Group B on September 5 and playoffs/final on September 6 as the decisive scheduled events, resolved by the official [Supercell event result](https://supercell.com/en/games/brawlstars/blog/esports/brawl-stars-challengers-finals-format--2/) (the ranking is only a proxy). |
| 09-01 04:36 | v82 | 🎃What will happen in October 2026? [ADD RESPONSES] :: winds of winter (next GoT book) release date announced | NO 0 | unfilled | model 0.01; edge -0.48; luna-medium price-blind says 0.01 vs answer 0.50 (edge -0.48 >= 0.3): Base rate is 0 confirmed release-date announcements in roughly 15 years (0/180 months); despite NYCC on October 8–11, the official GRRM Not a Blog, publications page, and appearances page show no date or relevant 2026 announcement, while publisher reporting rejects a fall release. |
| 09-01 04:36 | v82 | 🎃What will happen in October 2026? [ADD RESPONSES] :: china taiwan military conflict resulting in at least 5 d | NO 0 | unfilled | model 0.04; edge -0.46; luna-medium price-blind says 0.04 vs answer 0.50 (edge -0.46 >= 0.3): Base rate is 0 qualifying months in roughly 564 months since the 1979 end of cross-strait bombardment; the undefined “military conflict” and NA discretion add ambiguity and nudge the estimate upward, October 10 Taiwan National Day is the only salient scheduled date, the resolver has no named source (official MND activity reports are only a proxy), and the related Polymarket clash market is 7% but has materially different rules and no five-death threshold. |
| 09-01 04:36 | v82 | 🎃What will happen in October 2026? [ADD RESPONSES] :: usa bombs or missle strikes lebanon | NO 5 | open +1 | model 0.05; edge -0.45; luna-medium price-blind says 0.05 vs answer 0.50 (edge -0.45 >= 0.3): Base rate is roughly 0.3% of historical months since 1958 with direct U.S. strikes in Lebanon; ongoing regional conflict raises it modestly, but the rule is somewhat ambiguous and allows NA, the October 9 UN Security Council Lebanon session is the only scheduled relevant event, and the exact resolver would be an official DoD/CENTCOM announcement (with no matching witness markets found). |
| 09-01 04:36 | v82 | 🎃What will happen in October 2026? [ADD RESPONSES] :: A treasure from The Secret is found (12treasures.com) | NO 7 | open +1 | model 0.01; edge -0.49; luna-medium price-blind says 0.01 vs answer 0.50 (edge -0.49 >= 0.3): Base rate is 3 verified finds in roughly 528 months (0.57% monthly); the Oct. 9 Secret Hunt Community Getaway modestly raises activity, but no matching market witness exists, and the organizer’s broad NA clause plus unclear definition of “found” reduce confidence; resolver is 12treasures.com’s verified-find announcement. |
| 08-31 20:52 | v82 | Will nadya271 reach Codeforces International Master in or before 2027? | YES 72 | open +2 | model 0.99; edge +0.49; luna-medium price-blind says 0.99 vs market 0.50 (edge +0.49 >= 0.2): Base rate: International Master is 0.907% of all registered Codeforces users (and Master→IM progression is 33.10%); rules are unspecified but standard interpretation is attainment by the deadline, and official standings already show nadya271 at 2346, above the 2300 IM threshold ([Codeforces rank rules](https://codeforces.com/blog/entry/99660?locale=en&mobile=false), [standings](https://codeforces.com/contest/2119/standings/participant/213775242), [distribution](https://codeforces.com/blog/entry/151862?locale=en&mobile=true)), so scheduled rated contests before 2027-12-31 are immaterial, with the official Codeforces rating/profile as resolver and no matching witness market found. |
| 08-31 16:06 | v82 | NFL Week 12 - Prop Bets :: 35% or higher Kick Return rate | YES 0 | unfilled | model 0.99; edge +0.49; luna-medium price-blind says 0.99 vs answer 0.50 (edge +0.49 >= 0.3): Base rate is 32.8% in 2024 versus 74.5% in 2025, and this exact prop resolved YES in all seven observed 2025 instances; the undefined denominator/source creates a modest ambiguity, but the NFL’s unchanged kickoff incentives and 16-game slate from Nov. 25–30 make YES highly likely, settled by Manifold using official NFL box-score data ([NFL schedule](https://www.nfl.com/schedules/2026/by-week/week-12), [2025 return rate](https://www.nbcsports.com/nfl/profootballtalk/rumor-mill/news/nfl-had-74-5-percent-of-kickoffs-returned-in-2025-after-only-32-8-percent-in-2024)). |
| 08-31 16:06 | v82 | NFL Week 12 - Prop Bets :: For each letter of the alphabet, at least one touchdown is scored by a player th | YES 58 | open -6 | model 0.93; edge +0.43; luna-medium price-blind says 0.93 vs answer 0.50 (edge +0.43 >= 0.3): A 16-game slate and a strong historical hit rate make complete alphabet coverage highly likely, despite modest ambiguity around which credited touchdowns and name forms count. |
| 08-31 16:06 | v82 | NFL Week 12 - Prop Bets :: At least one missed kicking extra point attempt | YES 0 | unfilled | model 0.92; edge +0.42; luna-medium price-blind says 0.92 vs answer 0.50 (edge +0.42 >= 0.3): Recent NFL rates imply roughly 2–3 missed kicking extra points across a 16-game week, making at least one highly likely. |
| 08-31 16:06 | v82 | NFL Week 12 - Prop Bets :: Hip Drop Tackle penalty called during any game | NO 0 | unfilled | model 0.09; edge -0.41; luna-medium price-blind says 0.09 vs answer 0.50 (edge -0.41 >= 0.3): Base rate is 3 calls in 570 games (0.53% per game; about 8.2% across 16 games), with the 2026 rule still defining HDT under Rule 12-2-18; Week 12 runs Nov. 25–30 with 16 scheduled games, and the exact resolver is the NFL official gamebook/play-by-play (the stats are a nflfastR-based proxy), so despite minor ambiguity over no-play calls I estimate YES at 9% ([historical rates](https://www.nflpenalties.com/penalty/hip-drop-tackle?view=season), [2026 rulebook](https://operations.nfl.com/rules-officiating/2026-nfl-rulebook), [Week 12 schedule](https://www.nfl.com/schedules/2026/by-week/week-12)). |
| 08-31 15:28 | v82 | Will US translator jobs stay stable based on 2026 data (likely to be published Q1 2027)? | YES 43 | open +6 | model 0.84; edge +0.34; luna-medium price-blind says 0.84 vs market 0.50 (edge +0.34 >= 0.2): Base rate: CPS estimates were at least 89,000 in all six years 2020–2025 (91k, 93k, 94k, 101k, 92k, 97k), while the threshold is only 8.25% below 2025; the exact resolver is BLS CPS Table 11, expected to publish 2026 annual averages before the 2027-03-31 close, with modest uncertainty from sampling and the 2025 11-month comparability caveat. |
| 08-31 15:26 | v82 | GLM 5.4+ (Z.ai) release date :: Before 2026-11-01 | YES 57 | open +26 | model 0.86; edge +0.36; luna-medium price-blind says 0.86 vs answer 0.50 (edge +0.36 >= 0.3): Base rate: 5 of the last 6 GLM flagship upgrade intervals were 79 days or fewer; rules make any publicly accessible GLM 5.4+ count with little ambiguity; Z.ai’s rapid 5.1–5.3 cadence and API availability favor YES over sentiment; no specific scheduled event before November 1 is announced; the resolver is Z.ai’s official blog/docs/API availability ([GLM-5.3](https://z.ai/blog/glm-5.3), [release notes](https://docs.z.ai/release-notes/new-released)); the listed related markets provide no verified matching-rule price witness. |
| 08-31 15:21 | v82 | Claude Fable 5.1+ (Anthropic) release date :: Before 2026-09-11 | NO 54 | settled -54 | model 0.19; edge -0.31; luna-medium price-blind says 0.19 vs answer 0.50 (edge -0.31 >= 0.3): Anthropic’s unusually rapid Claude 5 cadence supports a meaningful near-term chance, but the only comparable market points to most Fable-release probability landing later in September and the scheduled Anthropic event is after the cutoff. |
| 08-31 15:21 | v82 | Claude Fable 5.1+ (Anthropic) release date :: Before 2026-11-01 | YES 100 | settled +52 | model 1.00; edge +0.47; luna-medium price-blind says 1.00 vs answer 0.53 (edge +0.47 >= 0.3): Base rate: 5/5 prior major Anthropic model launches were publicly accessible; Anthropic’s official newsroom records Claude Fable 5 released June 9, 2026 with API access, satisfying the rules before November 1, with no relevant scheduled event or matching-rule witness needed. |
| 08-31 15:21 | v82 | Claude Fable 5.1+ (Anthropic) release date :: Before 2026-10-11 | YES 60 | settled +42 | model 0.84; edge +0.34; luna-medium price-blind says 0.84 vs answer 0.50 (edge +0.34 >= 0.3): Base rate is roughly 60% for Anthropic shipping a major model in a 42-day window, but a matching related market is 86% before October and the Sep 23 Anthropic Founder House plus Sep 29 OpenAI DevDay create release opportunities; official Anthropic model docs/newsroom are the resolver, while current evidence is only rumor and gray-rollout proxy, so I discount modestly. |
| 08-31 15:19 | v82 | Kimi K3.1+ (Moonshot) release date :: Before 2026-11-01 | YES 65 | open +13 | model 0.86; edge +0.36; luna-medium price-blind says 0.86 vs answer 0.50 (edge +0.36 >= 0.3): Base rate is 3 recent Kimi releases in 2026 (K2.5, K2.6, K3), with roughly 2–3 month spacing; the rules allow K3.1/K4 and open API/product access, K3’s July release and reported arena testing support continuation beyond hype, no scheduled catalyst before November 1 is known, the resolver is Moonshot’s official Kimi model/API pages, and a closely matching market prices release by September 30 near 93%, warranting a modest downward adjustment for unconfirmed naming and access details. |
| 08-31 15:19 | v82 | Kimi K3.1+ (Moonshot) release date :: Before 2026-09-16 | YES 100 | open -28 | model 1.00; edge +0.50; luna-medium price-blind says 1.00 vs answer 0.50 (edge +0.50 >= 0.3): Base rate was roughly 0.5–0.7 for a qualifying release within 16 days given Moonshot’s recent ~2–3-month cadence, but the rules already resolve YES because Moonshot officially released Kimi K3 on 2026-07-16 and lists it as publicly available via its [official site](https://www.moonshot.ai/) and [API](https://platform.kimi.ai/docs/models), with no remaining calendar or resolver ambiguity. |
| 08-31 15:19 | v82 | Kimi K3.1+ (Moonshot) release date :: Before 2026-12-01 | YES 73 | open +16 | model 0.89; edge +0.39; luna-medium price-blind says 0.89 vs answer 0.50 (edge +0.39 >= 0.3): Base rate: 3/3 recent K-series releases occurred within 138 days of the prior major release; rules clearly count any publicly accessible K3.1+ version, current arena-testing reports support the narrative but are unconfirmed, no scheduled pre-December event is known, the resolver is Moonshot’s official Kimi blog/API model list, and mismatched witness markets imply roughly 0.93 by September and 0.98 by year-end, so I discount to 0.89 for naming and availability uncertainty. |
| 08-31 15:17 | v82 | GPT 5.7+ (OpenAI) release date :: Before 2026-11-21 | YES 11 | settled +5 | model 0.91; edge +0.41; luna-medium price-blind says 0.91 vs answer 0.50 (edge +0.41 >= 0.3): Recent GPT releases establish an 83% base rate, while DevDay timing and strong Astra-related market pricing raise the probability, offset by naming ambiguity and possible safety-driven access delays. |
| 08-31 15:17 | v82 | GPT 5.7+ (OpenAI) release date :: Before 2026-10-01 | YES 40 | settled +29 | model 0.80; edge +0.30; luna-medium price-blind says 0.80 vs answer 0.50 (edge +0.30 >= 0.3): Base rate: GPT-5.4, 5.5, and 5.6 launched in 126 days, with 49- and 77-day gaps, implying roughly 74% for another release within this 84-day window; the rules broadly accept GPT-5.7+, GPT-6, or equivalent publicly accessible branding, DevDay is scheduled for September 29, and Astra/GPT-6 witness markets imply strong but rule-mismatched odds, so I estimate 80%, resolved by OpenAI’s official release announcement or model catalog. |
| 08-31 15:15 | v82 | Muse Spark 1.3+ (Meta) release date :: Before 2026-11-16 | YES 56 | settled +42 | model 0.82; edge +0.32; luna-medium price-blind says 0.82 vs answer 0.50 (edge +0.32 >= 0.3): Base rate: Meta has released 3 Muse Spark versions in 145 days (about one every 48 days); rules qualify public API/product access and permit 1.4 or 2, Meta has promised a more powerful version later this year, Meta Connect is scheduled for September 23–24, and the exact resolver is Meta’s official AI blog/API model page, while the matching Polymarket witness prices an earlier September 30 cutoff near 46–52%, supporting but not determining a high probability. |
| 08-31 15:15 | v82 | Muse Spark 1.3+ (Meta) release date :: Before 2026-12-16 | YES 60 | settled +45 | model 0.84; edge +0.34; luna-medium price-blind says 0.84 vs answer 0.50 (edge +0.34 >= 0.3): Base rate: Meta has released two successive Muse Spark increments in 119 days (one every 59.5 days; 1.68 increments per 100 days), implying about 83% for another by the 107-day deadline; the rules clearly count the publicly accessible 1.2/API track, Meta has signaled larger models in development, Meta Connect on September 23–24 is the main scheduled catalyst, and the exact resolver is Meta’s official AI Research/developer model announcement pages [research](https://research.meta.ai/) [developer listing](https://ai.meta.com/llama/), with no matching-rule witness market available. |
| 08-31 15:15 | v82 | Muse Spark 1.3+ (Meta) release date :: Before 2026-11-01 | YES 56 | settled +42 | model 0.82; edge +0.32; luna-medium price-blind says 0.82 vs answer 0.50 (edge +0.32 >= 0.3): Base rate: 1 of 2 observed Muse Spark inter-release gaps was 62 days or less (50%), but releases accelerated from 92 to 27 days; rules are clear, Meta has publicly stated more capable models are in training, Meta Connect is scheduled for September 23–24, and the exact resolver is Meta’s official AI blog or Meta Model API availability page, with no verified matching-market witness used. |
| 08-31 15:15 | v82 | Muse Spark 1.3+ (Meta) release date :: Before 2026-12-01 | YES 60 | settled +45 | model 0.84; edge +0.34; luna-medium price-blind says 0.84 vs answer 0.50 (edge +0.34 >= 0.3): Base rate is 2/2 prior Muse Spark transitions within 93 days (100%, small sample); 1.2 is already publicly accessible, Meta says larger models are coming, Connect is scheduled for September 23–24, and the exact resolver is Meta’s official AI Research/Model API availability record, with no related-market prices used because matching rules could not be verified. |
| 08-31 15:10 | v82 | Grok 4.8+ (xAI) release date :: Before February 2027 | YES 91 | open +53 | model 0.87; edge +0.37; luna-medium price-blind says 0.87 vs answer 0.50 (edge +0.37 >= 0.3): Base rate is about 0.24 numbered/major Grok releases per month historically, with the recent 4.5→4.6 interval only 35 days; under the rules any publicly accessible 4.8+ qualifies, and Musk’s scheduled/explicit 3–4-week 4.7 target before the close plus xAI’s rapid cadence make a 4.8+ release by January 31 likely despite timeline slippage, with the official xAI News page and API model list as resolvers. |
| 08-31 15:10 | v82 | Grok 4.8+ (xAI) release date :: Before November 2026 | YES 100 | open +8 | model 1.00; edge +0.50; luna-medium price-blind says 1.00 vs answer 0.50 (edge +0.50 >= 0.3): Base rate is roughly 0.31 flagship/API model releases per month (4 from Jul 2025–Aug 2026); xAI’s official release notes record Grok 4.20 live on 2026-03-10, which satisfies >=4.8 and was API-accessible, so the market is effectively already YES; no calendar event or matching witness is needed, and the resolver is xAI’s release-notes/model catalog. |
| 08-31 15:07 | v82 | Gemini 3.8+ Flash (Google) release date :: Before 2026-10-11 | YES 16 | open +11 | model 0.86; edge +0.37; luna-medium price-blind says 0.86 vs answer 0.49 (edge +0.37 >= 0.3): Base rate: Google released two core Flash versions in the last 41 days, with the latest interval only 23 days; the rules require public access and exclude the current internal preview but otherwise are clear; hype is discounted because the cadence is documented; scheduled Google AI events run Sep 2–Oct 8; the resolver is Google’s official AI Developers model catalog/changelog and Gemini announcements; the closely matching Polymarket Sep-30 contract trades near 84%, adjusted slightly upward for internal testing and additional calendar catalysts. |
| 08-31 15:07 | v82 | Gemini 3.8+ Flash (Google) release date :: Before 2026-12-01 | YES 16 | open +11 | model 0.86; edge +0.37; luna-medium price-blind says 0.86 vs answer 0.49 (edge +0.37 >= 0.3): Historical base rate is 4 qualifying Gemini Flash releases in ~257 days (mean interval ~64 days; latest interval 23 days), and the 110-day window to Dec 1 therefore favors another release; the rules clearly include 3.9/4 Flash and public API or product access, while excluding Lite/Pro and restricted tests; Sep 2's scheduled Google Cloud Gemini Omni Flash event is relevant launch infrastructure but Omni is not itself a qualifying numbered Flash, and the exact resolver is Google's official Gemini API model list/endpoint, with no matching-rule witness from the linked markets used. |
| 08-31 15:07 | v82 | Gemini 3.8+ Flash (Google) release date :: Before 2026-11-01 | YES 21 | open +13 | model 0.94; edge +0.45; luna-medium price-blind says 0.94 vs answer 0.49 (edge +0.45 >= 0.3): Base rate is 2 of the last 2 Gemini Flash upgrades arriving within roughly two months; the rules clearly count any publicly accessible 3.8+ Flash via Google’s API/model list, no scheduled Google event before November 1 is known, the exact resolver is Google’s official Gemini API model catalog, and the matching Polymarket market prices October 31 at about 95%, modestly discounted for cadence uncertainty. |
| 08-31 15:07 | v82 | Gemini 3.8+ Flash (Google) release date :: Before 2026-09-21 | YES 17 | open +11 | model 0.87; edge +0.38; luna-medium price-blind says 0.87 vs answer 0.49 (edge +0.38 >= 0.3): Base rate is 2 of the last 2 Gemini Flash version transitions occurring within 39 days (the latest took 23 days); Google is iterating rapidly, the '+' allows a skip to 3.9/4 Flash, Sept. 15–18 Google AI events are the only relevant scheduled calendar window, and the resolver is Google's official Gemini API model/deprecation list (with leaks and related-market signals treated only as weak proxies). |
| 08-31 15:05 | v82 | Inkling 2+ (Thinking Machines) release date :: Before October 2026 | NO 27 | open +13 | model 0.08; edge -0.41; luna-medium price-blind says 0.08 vs answer 0.49 (edge -0.41 >= 0.3): Base rate is about 10% for a frontier lab to ship a publicly accessible numbered successor within 2.5 months; Inkling-Small’s rapid July 30 release modestly raises this, but it is not Inkling 2 under the rules, official news lists no scheduled successor before September 30, and the resolver is Thinking Machines’ official news/model page plus unrestricted API or weights availability, with no usable matching-rule witness markets found. |
| 08-31 15:02 | v82 | Qwen 3.9+ (Alibaba Qwen) release date :: Before 2026-12-01 | YES 15 | open +8 | model 0.84; edge +0.34; luna-medium price-blind says 0.84 vs answer 0.50 (edge +0.34 >= 0.3): Base rate: Qwen shipped numbered releases roughly every 56 days in 2026, implying about 0.81 probability of another within the 92-day window; Alibaba’s rapid Qwen3.5–3.8 cadence and Qwen3.8-Flash-Next preview of Qwen4 raise this to 0.84, with Qwen Conference Thailand on September 4 as a possible launch venue, while the resolver is an official Alibaba/Qwen API or public model listing (not the narrower Max-only prediction market, whose rules do not match) ([Qwen releases](https://github.com/QwenLM/Qwen3.5), [Alibaba Qwen3.8-Max announcement](https://www.alibabagroup.com/en-US/document-2021044032125272064), [Qwen Conference Thailand](https://www.techtalkthai.com/alibaba-cloud-qwen-conference-thailand-2026/)). |
| 08-31 15:02 | v82 | Grok 4.7+ (xAI) release date :: Before 2026-10-21 | YES 20 | settled +13 | model 0.91; edge +0.41; luna-medium price-blind says 0.91 vs answer 0.50 (edge +0.41 >= 0.3): Base rate: 2 of the last 2 xAI frontier-model gaps were under 70 days; rules count any publicly accessible 4.7+ variant, Musk’s 3–4-week target after the Aug. 12 Grok 4.6 release implies early September, no known mechanically relevant event before Oct. 21 is scheduled, and the resolver is xAI’s official news/API model listings ([announcement](https://x.ai/news/grok-4-6), [models](https://docs.x.ai/developers/models)), with no matching witness market identified. |
| 08-31 15:02 | v82 | Grok 4.7+ (xAI) release date :: Before 2026-10-11 | YES 18 | settled +12 | model 0.89; edge +0.39; luna-medium price-blind says 0.89 vs answer 0.50 (edge +0.39 >= 0.3): Base rate: xAI has released a qualifying public Grok update within roughly six months after each of its last four major releases (4/4), and Musk’s Aug. 12 estimate was 3–4 weeks for Grok 4.7; the rules count any accessible API/product release and higher version, with no known scheduled event before Oct. 11, while the exact resolver is xAI’s official model list/release notes (currently showing Grok 4.6), and no matching witness market was found ([xAI docs](https://docs.x.ai/developers/models), [release notes](https://docs.x.ai/developers/release-notes), [Musk timeline](https://x.com/elonmusk)). |
| 08-31 15:02 | v82 | Grok 4.7+ (xAI) release date :: Before 2026-09-21 | YES 14 | settled -14 | model 0.82; edge +0.32; luna-medium price-blind says 0.82 vs answer 0.50 (edge +0.32 >= 0.3): Base rate: 1 of xAI’s last 3 accessible flagship-version transitions occurred within 21 days; rules clearly count any publicly accessible Grok ≥4.7, Musk’s 3–4-week promise and the recent 4.5→4.6 cadence strongly raise the estimate, no known scheduled xAI event before Sep 21 mechanically changes it, and the resolver is xAI’s [official news](https://x.ai/news) plus [model catalog](https://docs.x.ai/developers/models) (the Coinbase market is excluded because its Sep 4 cutoff differs). |
| 08-31 15:02 | v82 | Grok 4.7+ (xAI) release date :: Before 2026-12-01 | YES 26 | settled +15 | model 1.00; edge +0.50; luna-medium price-blind says 1.00 vs answer 0.50 (edge +0.50 >= 0.3): Base rate is 6/6 recent named Grok releases becoming publicly/API accessible; xAI's official release notes already record Grok 4.20 live on 2026-03-10 (which exceeds 4.7), so the exact resolver is already YES, with no calendar or witness uncertainty material. [xAI release notes](https://docs.x.ai/developers/release-notes) |
| 08-31 14:40 | v82 | Will the U.S. Supreme Court grant cert or agree to hear arguments in the Texas DACA case by February 27, 2027? | NO 51 | open +11 | model 0.08; edge -0.25; luna-medium price-blind says 0.08 vs market 0.33 (edge -0.25 >= 0.2): Base rate is roughly 1–2% for cert petitions overall (about 20.7% for already-selected state petitions), but despite DACA’s importance, the pending Hanen ruling and likely Fifth Circuit process make review by 2/27/27 unlikely; no case is scheduled on the SCOTUS calendar before then, with the official SCOTUS docket as resolver and current status supported by [Informed Immigrant](https://www.informedimmigrant.com/resources/daca/daca-court/) and the [2026 Term calendar](https://www.supremecourt.gov/oral_arguments/2026TermCourtCalendar.pdf). |
| 08-31 13:55 | v82 | Will apple change Lake Ontario to lake America anywhere in the world? | YES 18 | settled +20 | model 0.62; edge +0.52; luna-medium price-blind says 0.62 vs market 0.10 (edge +0.52 >= 0.2): Base rate is 1/2 for comparable recent U.S.-directed geographic renamings adopted by Apple (Gulf of America yes, Denali no); Apple Maps currently still says Lake Ontario, but the official GNIS change and Gulf precedent support adoption before December 31, with no known scheduled catalyst, while missing rules and ambiguity over U.S.-only versus worldwide display pull the estimate toward 0.5; resolver: Apple Maps’ live label/place card, with Google Maps as a non-matching proxy witness. |
| 08-31 13:46 | v82 | Will Trump attempt to rename an ocean by the end of his term? | NO 30 | open +16 | model 0.24; edge -0.26; luna-medium price-blind says 0.24 vs market 0.50 (edge -0.26 >= 0.2): Trump’s recent renaming pattern and explicit Atlantic/Pacific comments raise the probability above the near-zero historical base rate, but no concrete ocean-renaming action exists and the rule’s ocean-versus-gulf ambiguity remains material. |
| 08-31 08:58 | v82 | Openai Astra publically available before September 7th 2026 | NO 26 | settled -26 | model 0.12; edge -0.38; luna-medium price-blind says 0.12 vs market 0.50 (edge -0.38 >= 0.2): Base rate: I estimate only ~2% for a major frontier-model release in any random seven-day window; Astra is explicitly still upcoming and its release was slowed for cyber-safety work [Axios](https://www.axios.com/2026/08/07/openai-astra-model-delay-cybersecurity-risks), while OpenAI’s scheduled September 2–3 events are educational rather than launch events [OpenAI Academy](https://academy.openai.com/public/events); the resolver would be an official OpenAI product/announcement page, but missing rules create ambiguity over partner access versus genuine public availability, so I modestly pull the estimate toward 0.5 despite release rumors. |
| 08-31 04:58 | v82 | Will Waymo expand it's LA service area by the end of 2026? | YES 0 | unfilled | model 0.84; edge +0.34; luna-medium price-blind says 0.84 vs market 0.50 (edge +0.34 >= 0.2): Base rate: Waymo announced 3 LA service-area expansions in the prior ~16 months (~19% monthly, ~53% over four months); CPUC approved a broader LA territory on Aug. 14, 2026 with gradual rollout planned, making a further expansion by Dec. 31 likely, though the LA/city-versus-metro wording is mildly ambiguous; the resolver is Waymo’s official service-area map/app and no fixed LA rollout date is scheduled. [Waymo announcements](https://support.google.com/waymo/announcements/12766613?hl=en) [CPUC approval coverage](https://www.cbsnews.com/sacramento/news/waymo-approved-sacramento-san-diego-expand-los-angeles-bay-area/) |
| 08-30 22:32 | v82 | Best Climber classification winner in la Vuelta a España :: Leknessund | NO 14 | settled +4 | model 0.08; edge -0.17; luna-medium price-blind says 0.08 vs answer 0.25 (edge -0.17 >= 0.1): Base rate for one rider among roughly 20 realistic Vuelta KOM contenders is about 5%, and despite Leknessund’s stage-7 win he is now third on 19 points behind Buitrago (29) and Mas (24) with mountain stages through September 12 remaining; the rules clearly require the final official La Vuelta rankings, while matching BetMGM odds of 19.00 imply about 5.3%, so I modestly adjust upward for his proven breakaway profile. [Official rankings](https://www.lavuelta.es/en/rankings) [current standings](https://cyclinguptodate.com/cycling/vuelta-a-espana-2026-classifications-update-stage-9-enric-mas-strengthens-red-jersey-as-oscar-onley-closes-on-podium) [market witness](https://www.betmgm.com/en/sports/events/vuelta-a-espana-2026-stage-9-19882094) |
43 trades, 1,505 staked, realised +294, open +307 if held / +187 if sold, 6 unfilled.
In the traces on hand the model priced 0
answers across 0 markets and the module acted on
0 of them (0%). The journal holds
43 trades for 1,505 mana. 18 of them settled for +294.
An estimate the model declines to give is treated as no opinion and never
as a cheap one.
Traces are diagnostic and rotate; the trading journal is the book above.
Built 2026-09-22 18:05Z.